Why Robots Won’t Be Taking Over the World Anytime Soon

The woman waved her clipboard as the crowds grew around her. Flapping away on the exhibition floor at the ExCeL conference center in east London, she sounded irritated: a key person was missing, no one was lining up in an orderly fashion and it was well past time to start. As it turns out, organizing a robot parade is as difficult as herding cats.

This was ICRA 2023—to the uninitiated, the Institute of Electrical and Electronics Engineers (IEEE) International Conference on Robotics and Automation—a major robotics conference featuring luminaries from Boston Dynamics founder Marc Raibert to cybernetics art critic Jasia Reichardt and body-double avatar maker Hiroshi Ishiguro, as well as startups, students, and inventors showing off their creations. Though everyone visiting ICRA had days to wander the booths looking at every robot set to take part in this parade, the appeal of seeing them stomp about en masse had gathered hundreds of attendees.

The organizer with the clipboard shouted the robot wranglers into position, found the missing luminary to grand marshal the parade and cleared the aisles for the procession. Just as the show was about to begin, a random man carrying a box stepped up to a robot and kicked it over before scurrying away, his motivation obscure; the robot was quickly set right.

The parade began to march. Many looked like smaller versions of Boston Dynamics’ Big Dog—the four-legged one with backwards knees you’ve probably seen on YouTube—but those were joined by wheeled tanks waving jointed arms, tiny toy-like creations and a large, flat box like a mobile fax machine. A bat-like drone fluttered overhead for a few minutes. All of the humanoid robots used wheels rather than legs. The box-wielding robot kicker aside, there was no drama, nor any surprises during the parade. With a call of “watch out, robots coming through,” they trundled up one aisle of the show floor before returning down another, lined by people snapping photos and recording video (myself included).

Beyond bipedal omissions, there was something else to note: every parading robot was followed by at least one human, most holding what looked like a games console controller in their hands. And that’s because none of these robots were smart enough to walk up and down a hall full of other robots and people without being directed like a radio-controlled toy car. That’s the often-unrealized secret of most modern robots: we’re afraid of them taking over the world, but most of them can’t do anything beyond walking a few steps without our instructions.

Robots aren’t the future—they’re here and now and have been for decades, though what they’re capable of (and not) may surprise you. Robots dominate production lines, assist in surgery, and vacuum our floors. What each of these examples has in common is specificity of use and a practical design to match, and that means their shapes don’t resemble us humans. The robots building cars are giant jointed limbs that can lift and manipulate into place components both heavy and delicate. The da Vinci line of surgical robots wield multiple instruments in their precision arms, but are directly controlled by surgeons. And Roombas are small, flat suction machines, handily rideable by cats—and the closest most of us will come to a robo-butler.

Carefully developed for specific uses, these designs make sense. A humanoid robot pushing a hoover around is ridiculous and inefficient. Why not simply automate the vacuum? Yet the dream of an android, a humanoid robot, persists. We can’t stop trying to build a do-it-all, mechanized person. And we have dreamed of it for centuries—Leonardo da Vinci himself in 1495 created a metal knight that could wave its arms and sit down to entertain at an Italian noble’s parties.

More than 500 years later, Elon Musk stepped onstage to announce his take on humanoid robots, though not with a working demo but with dancers dressed like robots; it looked like a student theatre performance. We just can’t help it. To understand why we want robots that reflect our image, you’d need to study philosophy, religion and psychology—and you’d still be left scratching your meat-based head. Here, we’ll just accept it as a persistent fact of humanity, and move on to how to make the darn things. Because that’s a tough enough challenge.

Anyone who has watched an infant shift into their toddler years with their first wobbling steps knows that walking isn’t easy. Unlike newborn deer or giraffes, up and running away from predators mere moments after birth, it takes us humans years to perfect this core movement. And it’s not just walking: try not to scream “that’s not how you do it” when sitting with a toddler slamming the incorrect block into a shape sorter or opening the flap of a book by tearing it apart, and it’s clear that precision doesn’t come easy to fumbling humans. Learning more complicated movements—swinging a bat to hit an arcing ball, cutting a carrot without slicing fingers, pulling a bow across strings to make pleasing sounds—requires careful study and practice.

And that’s despite our massive brains and well-evolved bodies. Imagine starting from scratch. The very first robots, if you can even call them that, arrived in the 1920s, with metal bodies and a few clunking motions, though some eventually included speech and one even smoked a cigarette. We quickly put them to work, or at least we did those robotic jointed arms that could manipulate components. These started working in 1961 in a General Motors (GM) car factory, and a few years later, artificial intelligence was added to a rolling robot in California so that it could navigate on its own—it probably says a lot about humans that we started with bodies rather than brains. In the 2000s, Boston Dynamics designed a robot soldier of sorts for the US military, and it all culminated in Japan with what was perhaps the first true humanoid android, Honda’s Asimo.

Bipedal robots have done parkour (Boston Dynamics again), chatted to UN diplomats (Hanson Robotics’ Sophia), and acted in theatre (Engineered Art’s RoboThespian). That’s on a good day—androids are as often laughable for their obvious flaws, with supercut videos of robotic gymnasts falling on their faces just as popular online as those where they land their backflips.

But there’s something lacking in robots: intelligence. We can program robots to do anything, but we still have to tell them what to do—indeed, the latest models of humanoid robots are avatars that are remote-controlled from a distance by a human operator. And though sci-fi films spark dystopian nightmares and (as with AI in the previous chapter) fears of lost jobs, the reality of automated androids is closer to the Knightscope K5, a security robot that rolled into a fountain at a Washington, DC office in 2017 and drowned. Rest in peace, Steve the robo-guard. We’ll never forget you.

*

A long 89 years before Steve’s damp demise, William Richards had a problem. A very well-known public figure—often reported as being the Duke of York (later King George VI), though others say it was merely a renowned scientist—was booked to speak at the opening of the 1928 Exhibition of the Society of Model Engineers that Richards was organizing in London, but cancelled at the last minute. Who could replace this lost-to-history luminary?

Richards had a solution: “make a man of tin.” And that he did—well, he used aluminium. Eric, as the metal speechmaker was inexplicably named, wasn’t exactly a robot as it couldn’t walk but stood where its handlers left it. But its head and arms moved, and its eyes lit up—as did its mouth, with 35,000 volts of electricity sending blue sparks flying from its teeth to indicate anger. It’s not clear exactly how Eric worked, but two people were required for its physical operation, suggesting it was more a man-handled, programmable automaton than a robot.

Where this metal man did show something approaching intelligence, or at least automation, was in speech. Eric could stand, wave and turn its head on verbal command, but also answer basic questions, like a giant metal voice assistant. One journalist ‘interviewing’ the robot asked if it was married. “But before he replied I knew it was a foolish question. He looked too happy,” writes the journalist, because this was the 1930s. Eric replied that it was “not likely” to be married, “because I am too young. I am only 18 months old.” Eric was “so amiable” about all the questions that the journalist queried if it could ever lose its temper: “He replied by staring at me with a fearsome look and spitting a shower of sparks at me.”

The voice magic was done through radio, but at the time Richards wouldn’t reveal exactly how the question-and-answer sessions worked. “All I can say,” he said, “is that in constructing the ‘man’ we have used the most advanced methods of radio control…using some of the [Marconi company’s] patents. I can assure you, however, that his speech is produced neither by phonograph record or talking film.”

Massive, limbed radio though Eric may have been, that was novel enough for 1928. Eric’s initial four-minute address at the Exhibition of the Society of Model Engineers was a rousing success, so the metal man toured the UK, even deputizing for the lord mayor in Newcastle, before heading stateside. The success of Eric and its contemporaneous robot rivals inspired others, including Elektro and its dog Sparko, which were created for the 1939 World’s Fair, though again, purely as entertainment.

But it was believed even then that human-shaped machines had the potential to take on jobs normally held by people: in 1933, a newspaper reported rumors that London County Council had ordered a humanoid robot from Eric’s makers to photograph cars running traffic lights, though there’s no evidence it ever took on that role. Westinghouse Electric Corporation’s Televox, a similarly basic metal man that could respond to voice commands, stood in for watchmen at reservoirs in Washington, DC, perhaps marking the first jobs lost to robots, albeit just for the one shift. Given the robot could do little other than speak and stand, if a wrongdoer attempted to sneak past during the stunt, Televox couldn’t have put up much resistance beyond a verbal ticking off.

But before Eric could get a job outside the world of entertainment, it disappeared—possibly for scrap—sometime after 1936, as that’s when mentions in newspapers trail off. Several decades later, in 2016, the Science Museum in Kensington, London, rebuilt a version of Eric to star in its exhibition on robots. Because of Eric’s disappearance, and Richards’ reluctance to share its workings, Science Museum curator Ben Russell and roboticist Giles Walker had to make it up as they went, re-engineering Eric using modern components. To reproduce the voice in the modern Eric, Russell tells me he recorded snippets of his own speech on his phone for playback later.

Though Eric and its fellow metal friends were far from capable, reporters of the time saw potential. “His actions may be limited at present, but so were the movements of the first cars—only 35 years ago,” noted the Croydon Times in 1936. “If Eric can do all he can today, what will he do in 50 years time?”

A lot, as it turns out. While Eric was little more than an entertainment box built by a clever showman, by the mid-1980s, robots were starting to learn how to walk on their own—though it took some effort figuring out how to make legs work, and also to teach the machines how to think (or something like it). But as it happens, we had managed to put them to work well before then.

__________________________________

Excerpted from The Long History of the Future: Why Tomorrow’s Technology Still Isn’t Here by Nicole Kobie. Used with the permission of the publisher, Bloomsbury Sigma, an imprint of Bloomsbury Publishing. Copyright © 2024 by Nicole Kobie.